Abstract：In order to improve the precision of small scale target detection in remote sensing image and realize real-time detection, a remote sensing image target detection algorithm based on deep learning is proposed. Firstly, a large-scale remote sensing image dataset was constructed for model training and testing. Secondly, based on the original Single Shot multibox Detector network (SSD) model, the shallow feature fusion module, shallow feature enhancement module and deep feature enhance module were designed. Finally, the focal loss function was introduced into the training strategy to solve the problem of the imbalance of positive and negative samples in the training process, and the experiment was carried out on the remote sensing image dataset。Experiments on high-resolution remote sensing image data sets show that the mean Average Precision (mAP) of improve-SSD algorithm achieves 77.95%, which is 3.19 percentage points higher than SSD, at the speed of 33.8 frame/s. On the extended experiment, improve-SSD performance is better than SSD for the detection of fuzzy targets in high-resolution remote sensing images. Experimental results show that the improve-SSD algorithm proposed in this paper can effectively improve the accuracy of remote sensing image target detection.
